6.5.1 Introduction

For more than one and a half decades, solar irradiance (both bolometric and at different wavelengths) has been observed from different satellites. It has been shown that total solar and UV irradiances vary in parallel with the solar magnetic activity cycle, being higher during maximum solar activity conditions (e.g. Hudson, 1988). The long-term irradiance variations over the solar cycle are attributed to the changing emission of bright faculae and the magnetic network (e.g. Lean, 1987). The short-term irradiance changes (on the time scale of solar rotation) are directly related to the effect of active regions as they evolve and move across the solar disk (Lean, 1987) and to a lesser extent to the bright magnetic network (Pap, Marquette and Donnelly, 1991). Although considerable information exists on solar irradiance variability, the underlying physical mechanisms are not well understood as yet (see summary by Pap and White, 1994). Because of the lack of quantitative physical models of the solar radiative output, the current irradiance ``models'' have been developed with linear regression analysis, which provides only a general information about the observed irradiance changes. While there is a reasonably good agreement between measured total solar and UV irradiances and their regression estimates calculated from observations of dark sunspots, bright faculae and the magnetic network during the declining portion of solar cycle 21 and the rise of solar cycle 22, there is a breakdown between the observed and estimated irradiance values during maximum and minimum solar activity conditions.

It has been shown that the calculated total irradiance values underestimate the observed changes at the maximum of solar cycle 21 (Foukal and Lean, 1988; Willson and Hudson, 1991) and solar cycle 22 (Pap et al., 1994; Brandt, Stix and Weinhardt, 1994). In the case of UV irradiance, the calculated irradiance values cannot account for the slow decrease in UV irradiance during solar minimum activity conditions (Donnelly, 1989; Barth et al., 1990; Pap, London and Rottman, 1991). Furthermore, the ``remaining part'' of the irradiance power spectra after removing the effect of sunspots and bright magnetic features, including faculae, plages and the magnetic network, has peaks at periods of 9, 13.5, 27 and 300 days (Fröhlich and Pap, 1989; Pap and Fröhlich, 1992; Pap, 1992), and the amplitude of this residual variability depends on the phase of the solar cycle (Pap and Fröhlich, 1992).

It has not been clarified yet whether these discrepancies between the observed and calculated irradiance data are simply related to measuring uncertainties (in both irradiance and ground-based surrogates) or they characterize a real solar variability that may be related to global effects, such as large scale motions (Ribes, Mein and Mangeney, 1985), radius changes (Ulrich and Bertello, 1995) and/or photospheric temperature changes (Kuhn, Libbrecht and Dicke, 1988). Considering that the radiative output of the Sun is the main driver of the physical processes in the Earth's atmosphere and its climate system, this problem has become one of the most important issues in solar-terrestrial physics.

The main objective of this paper is to (1) separate the observational uncertainties and physical fluctuations in total solar irradiance and (2) determine the amplitude of the real fluctuations in the examined time series over the solar cycle. In order to interpret the results gained from various types of statistical analyses, it is essential to understand and evaluate the uncertainties in the data. Identification of random fluctuations in the data sets provide an important clue to estimate the signal-to-noise ratio and to clarify whether the observed small residual variability in solar irradiance is simply determined by a noise component in the data or whether it clearly represents a real solar variability. The corresponding results on the distribution of daily fluctuations (related to both instrumental and solar noise) of solar irradiance and its ground-based surrogates are described in this paper.